An Algorithm Study For Fingerprint Classification And Matching Based On Singular Points

Along with development of society, traditional security systems which based mainly on token or password become fragile and cannot meet the demand of modern security system. Which makes people research for new security system that is safer and more difficult to fabricate than the traditional security system. Fingerprint recognition technique that is safer and has more feasibility was appeared under this circumstance. Fingerprint recognition technique can use our body's physical characters to make itself becoming a safer security system. Modern statistics has proved that fingerprint is unique and steady, that means, one's fingerprint is different from all others and will not change all his life. As one kind of biometric recognition technique, fingerprint recognition technique is widely used in forensic, police and other security system. Recently, as the development of micro –computer technique and embed developing system technique, automated fingerprint identification system (AFIS) has made a great progress. This article is meanly emphasizes on the fingerprint image processing technique that is one of the key techniques of automated fingerprint identification system. The target of author is make an automated fingerprint identification system in personal computer. The post-processing of fingerprint identification algorithm which include detection of singular points, fingerprint classification, fingerprint matching, fingerprint identification and the construction of local structure is concentrated by the author. The efficiency and effect of the algorithm must be considered when it was loaded into embedded system based DSP chip. This article put the emphases on the question described as follows: 1) The effectiveness and correctness of detection of singular points for different quality fingerprint images is very important, so we make two methods to detect singular points of fingerprint. One is based on Poincare index, the other is based on direction probability. It is more effectiveness and correctness for the detection of singular points by using the two methods combinational. 2) After the detection of singular points, we get the number, position and type of these singular points. We can get the result of fingerprint classification by count the number of core points and calculate the relative position of core point and delta point. The algorithm of classification is very simple and could be realized in embedded system. 3) This article introduces a local structure matching algorithm based on singular points. The local structure is constructed by a minutia and its 4-nearest neighborhood minutiae. After matched all local structures of two fingerprint images, system find out the structure pair which matching score is most high and make the structure's central point as the reference point of fingerprint image. Aligns the fingerprint image based on reference point, and matches minutiae of two the fingerprint images. System makes a decision whether the two fingerprints are come from same finger by count the number of matching minutiae pairs of the two fingerprints.